add a test case with the model of sample_multiple_hybrid_dims and m…

…odify `sample` model

model_repository/sample/1/model.py

@@ -17,7 +17,7 @@ def execute(self, requests):
         responses = [None for _ in requests]
         for idx, request in enumerate(requests):
             current_add_value = int(json.loads(request.parameters()).get("add", 0))
-            in_tensor = [item.as_numpy() + current_add_value for item in request.inputs() if item.name() == "model_in"]
+            in_tensor = [item.as_numpy() + current_add_value for item in request.inputs() if "model_in" in item.name()]
             out_tensor = [
                 pb_utils.Tensor(output_name, x.astype(output_dtype))
                 for x, output_name, output_dtype in zip(in_tensor, self.output_name_list, self.output_dtype_list)

model_repository/sample_multiple_hybrid_dims/1/model.py

@@ -0,0 +1,45 @@
+import json
+
+import numpy as np
+import triton_python_backend_utils as pb_utils
+
+
+class TritonPythonModel:
+    def initialize(self, args):
+        self.model_config = model_config = json.loads(args["model_config"])
+        output_configs = model_config["output"]
+
+        self.output_name_list = [
+            output_config["name"] for output_config in output_configs
+        ]
+        self.output_dtype_list = [
+            pb_utils.triton_string_to_numpy(output_config["data_type"])
+            for output_config in output_configs
+        ]
+
+    def execute(self, requests):
+        responses = [None for _ in requests]
+        for idx, request in enumerate(requests):
+            current_add_value = int(json.loads(request.parameters()).get("add", 0))
+            in_tensor = [
+                item.as_numpy() + current_add_value
+                for item in request.inputs()
+                if "model_in" in item.name()
+            ]
+
+            out_tensor = [
+                pb_utils.Tensor(output_name, x.astype(output_dtype))
+                for x, output_name, output_dtype in zip(
+                    in_tensor, self.output_name_list, self.output_dtype_list
+                )
+            ]
+            inference_response = pb_utils.InferenceResponse(output_tensors=out_tensor)
+            out_tensor.append(
+                pb_utils.Tensor(
+                    "model_out2",
+                    np.array([current_add_value], dtype=self.output_dtype_list[1]),
+                )
+            )
+
+            responses[idx] = inference_response
+        return responses

model_repository/sample_multiple_hybrid_dims/config.pbtxt

@@ -0,0 +1,63 @@
+name: "sample_multiple_hybrid_dims"
+backend: "python"
+max_batch_size: 2
+
+parameters [
+  {
+    key: "add",
+    value: { string_value: "0" }
+  }
+]
+
+input [
+{
+    name: "model_in0"
+    data_type: TYPE_FP32
+    dims: [ -1 ]
+},
+{
+    name: "model_in1"
+    data_type: TYPE_FP32
+    dims: [ -1 ]
+}
+]
+
+output [
+{
+    name: "model_out0"
+    data_type: TYPE_FP32
+    dims: [ -1 ]
+},
+{
+    name: "model_out1"
+    data_type: TYPE_FP32
+    dims: [ -1 ]
+},
+{
+    name: "model_out2"
+    data_type: TYPE_FP32
+    dims: [ 1 ]
+}
+]
+
+instance_group [{ kind: KIND_CPU, count: 1 }]
+
+model_warmup {
+  name: "RandomSampleInput"
+  batch_size: 1
+  inputs [{
+      key: "model_in0"
+      value: {
+        data_type: TYPE_FP32
+        dims: [ 10 ]
+        random_data: true
+      }
+   }, {
+      key: "model_in1"
+      value: {
+        data_type: TYPE_FP32
+        dims: [ 10 ]
+        zero_data: true
+      }
+   }]
+}

tests/test_model_call.py

@@ -69,3 +69,22 @@ def test_reload_model_spec(config):
     sample = np.random.rand(8, 100).astype(np.float32)
     result = client(sample)
     assert np.isclose(result, sample).all()
+
+
+def test_with_multiple_hybrid_dims(config):
+    client = get_client(*config, model_name="sample_multiple_hybrid_dims")
+
+    samples = [np.random.rand(1, 100).astype(np.float32) for _ in range(2)]
+    ADD_VALUE = 1
+    result = client(
+        {
+            client.default_model_spec.model_input[0].name: samples[0],
+            client.default_model_spec.model_input[1].name: samples[1],
+        },
+        parameters={"add": f"{ADD_VALUE}"},
+    )
+
+    assert np.isclose(result[0], samples[0][0] + ADD_VALUE).all()
+    assert np.isclose(result[1], samples[1][0] + ADD_VALUE).all()
+
+    assert np.isclose(result[2], ADD_VALUE).all()